All structures such as bridges, buildings and dams are deteriorated by environmental conditions like corrosion, earthquake and traffic after they are built. Therefore, vibration-based damage identification (VBDI) techniques have attracted attention to verify the safety and functionality of structures. Conventional damage identification techniques such as non-destructive evaluation (NDE) methods have many obstacles and are not practical to be implemented in order to detect damage. One of the main obstacles is that the conventional methods are focused on local structural damage and easily affected by measurement noise. Moreover, there are several uncertainties that restrict the successful application of the damage detection. On the other hand, vibration-based techniques can be used when the global vibration characteristics is available. To do this, it is required to perform modal identification, perform model updating, and detect local changes in a structure. The objectives of this study are: (a) to develop vibration-based techniques for modal identification by output-only information through the experimental and numerical study of bridges and frame structures; and (b) to study the VBDI techniques to perform model updating and damage detection using the changes in the dynamic characteristics of structures and determine their performance in practical structures. Two existing bridges, a Pre-stressed Concrete Bridge (PSCB) and a Steel Box Bridge (STB) tested earlier were used here as case studies and the data from the vibration test were used to verify modal identification using output-only information. In this study, two output-only methods, Frequency Domain Decomposition (FDD) and Stochastic Subspace Identification (SSI) were studied to compare the results from the vibration test. In addition to this, laboratory experiments on three-story steel frame were carried out to generate additional data for testing the system identification and model updating methods. Further, various damage scenarios were created in the frame to obtain the vibration signals corresponding to such conditions. The vibration data from the undamaged and damaged frame were used to study the performance of existing VBDI methods. Based on the above study, it is concluded that both FDD and SSI techniques provide accurate results for modal identification in the case of studied structures. FDD is relatively simpler, but it may miss some modal information when two modes are closer. However, in the cases studied here, the modes are quite apart from each other. Among many available VBDI methods, the matrix update method performs better than others when the measurement noise is small. Otherwise, data-driven models such as those based on Genetic Algorithm should be used.